Computational Discovery of Pharmacological Chaperones to Rectify Protein Misfolding Using a Novel Support Vector Machine Classifier
By Hajira Fuad
Ever since I was a kid, I’ve been both fascinated and frightened by that full range of malicious, deadly human diseases that have no cure. The possibility of one’s body turning against itself, waging war on its own cells and disrupting the complex biological processes that keep us healthy terrified me. I found the best way to confront my fears was to simply learn what causes debilitating diseases like cancer and Alzheimer’s disease what goes wrong in our body to cause these horrible maladies, and why. Somehow, learning about the purely technical, scientific aspects behind the pathogenesis of these diseases helped to erode my sense of powerlessness. I became hopeful and naturally progressed to thinking about cures. I indulged my growing curiosity by getting my hands dirty and reading esoteric abstracts, which, with lots of help from Google, I gradually became able to understand … Pharmacological chaperones are orally administered small molecules that, when bound to a misfolded protein, revert the misfolding process. To discover pharmacological chaperones for specific protein targets, knowledge of the 3D structure of the protein is required to identify exosites for the chaperone to bind to. Even then, most misfolded proteins do not possess natural binding sites. This project aims to find the structural analogues of ligands of misfolded proteins that can function as pharmacological chaperones. Using Java, I developed a classifier based on the support vector machine learning model to predict the structural similarity between two molecules using 2D molecular descriptors that function as similarity metrics.